This is a preprint of an article whose final and definitive form will be published in The Philosophical Quarterly. A shorter version of this article was presented at the Pacific Division of the APA, San Francisco 2007.

Sea Battle Semantics

Berit Brogaard May 2, 2007

The assumption that the future is open makes well-known problems for traditional semantics. We have a strong intuition to the effect that today’s occurrence of the sentence ‘there will be a sea battle tomorrow’, while truthvalueless today, will have a determinate truth-value by tomorrow night. Yet given traditional semantics, sentences that are truth-valueless now cannot later “become” true. Relativistic semantics supposedly does a better job accommodating our intuitions about future contingents than does non-relativistic semantics. I will argue, however, that our intuitions about future contingents cannot by themselves motivate such a paradigm shift, for, initial appearances to the contrary, standard non-relativistic semantics (plus an account of truth-value gaps) can accommodate both of our intuitions about future contingents.

I. The Problem of Future Contingents Imagine that at some time t1 I assert the following sentence: 1

(1)

There will be a sea battle tomorrow

Imagine furthermore that nothing about the present state of the world fully determines whether a sea battle will take place tomorrow. Relative to one possible future a sea battle will take place; relative to another a sea battle will not take place. My assertion of (1), it seems, is then neither determinately true nor determinately false.

1

I shall use ‘time’ to refer to what MacFarlane calls a ‘moment’. See ‘Future Contingents and Relative Truth’, The Philosophical Quarterly 53 (2003), pp. 321-36.

1

We can account for truth-value gaps in a variety of ways; one is to treat them in the way suggested by Arthur Prior. 2 Prior’s approach to truth-value gaps is a kind of supervaluation: a future tensed claim is true iff it expresses a proposition that is true relative to all possible continuations of the current state of the world, false iff it expresses a proposition that is false relative to all possible continuations, and neither true nor false iff it expresses a proposition that is true relative to some possible continuations, and false relative to others. One virtue of Prior’s approach is that it allows us to be logical classicists: all instances of Excluded Middle come out true. My current assertion of ‘either there will be a sea battle tomorrow, or there will not be a sea battle tomorrow’, for example, is true relative to all possible continuations of the current state of the world. Supervaluation captures our current intuitions about the truth-value of (1) with respect to my current context. But it has trouble accommodating tomorrow’s intuitions about (1) as uttered by me today. When we are actually standing in the harbor in the midst of a sea battle, it is tempting to think that my earlier assertion of (1) was true. However, standard semantics seems unable to accommodate this intuition. On standard semantic theory, a context of use is a sequence of parameters, which, following David Kaplan, 3 include a speaker, an addressee, a world, a time, and a location (of course, if the universe is truly indeterministic, there will not be a unique world but a set of worlds). The context of use plays two distinct roles: it fixes the semantic values of indexicals, and it determines an index with respect to which the proposition expressed by the sentence is evaluated for truth. 4 Which parameters of the context go into the index will depend on which index-shifting operators the language contains. Assuming that the language contains only modal 2

Time and Modality (Oxford: Oxford University Press 1957). D. Kaplan, “Demonstratives”, in J. Almog, H. Wettstein, and J. Perry (eds), Themes from Kaplan (New York: Oxford University Press, 1989), pp. 233-249. 4 For ease of exposition I shall call the semantic value of sentences ‘propositions’, even if this terminology is not entirely happy. See Jeffrey C. King, ‘Tense, Modality, and Semantic Values’, in John Hawthorne and Dean Zimmerman (eds.), Philosophical Perspectives 17 (2003): Language and Philosophical Linguistics, pp. 195-246. 3

2

operators and tense operators and the universe is truly indeterministic, indices will be pairs of a set of worlds and a time. 5 Index-shifting operators shift the parameters of the indices of evaluation. ‘It is possible that’, for example, shifts the world feature of the index, and ‘it has been that’ shifts the time feature of the index. In the case of ‘it has been that there are dinosaurs’, for example, ‘it has been that’ shifts the time feature of the index determined by the context of use to a time t such that t is earlier than the time of speech. Given standard semantics, then, (1), as uttered by me at t1, is true simpliciter iff the proposition expressed by (1) at my current context of use is true at the index determined by that context. Any potential future speech situation at which my assertion at t1 is evaluated is irrelevant to whether or not the assertion is true. So, since my assertion of (1) at t1 is neither determinately true nor determinately false relative to my context of use at t1, it cannot later turn out to be true. But later when we are standing in the harbor in the midst of a sea battle, it surely seems that my earlier assertion was true at the time of utternace. This is the problem of future contingents.

II. Double-Time Reference Semantics John MacFarlane has recently developed a new relativistic semantics that can account for both of our intuitions about future contingents. 6 MacFarlane’s semantics, which is based on Nuel Belnap

5

Other possible parameters include a location parameter and a standard of precision parameter. I shall ignore these parameters in what follows. The assumption that there are tense operators in English has come under recent attack. See Jeffrey C. King, ‘Tense, Modality, and Semantic Values’. Here I shall assume that it is unproblematic. Nothing of substance hangs on this. 6 ‘Future Contingents and Relative Truth’, and ‘Truth in the Garden of Forking Paths’, manuscript. For other defenses of relativism see e.g. MacFarlane, ‘Epistemic Modalities and Relative Truth’, manuscript, “The Assessment Sensitivity of Knowledge Attributions”, in Tamar Szabó Gendler and John Hawthorne (eds.), Oxford Studies in Epistemology 1 (Oxford: Oxford University Press, 2005), pp. 197-233, Max Kölbel, Truth Without Objectivity (Routlegdge: London, 2002), ‘Faultless Disagreement’, Proceedings of the Aristotelian Society 104 (2003), pp. 53-73, Mark Richard, ‘Contextualism and Relativism’, Philosophical Studies 119 (2004), pp. 215-42, Lasersohn, ‘Context Dependence, Disagreement, and Predicates of Personal Taste’, Linguistics and Philosophy 28 (2005), pp. 643-86, at p. 655, A. Egan, J. Hawthorne and B. Weatherson, ‘Epistemic Modals in Context’, in G. Preyer and G. Peter (eds.), Contextualism in Philosophy (Oxford: Oxford University Press, 2004), Egan, ‘Epistemic Modals, Relativism and Assertion.’ URL http://www.geocities/might.doc, manuscript, and Weatherson,

3

and Mitchell Green’s double-time reference semantics, 7 rejects the standard assumption that sentence truth is relative only to a context of use. Sentence truth, on MacFarlane’s view, is relative to both a context of use and a context of assessment. Using the above terminology rather than MacFarlane’s, a context of use and a context of assessment together determine a set of worlds that run through the time of the context of use and the time of the context of assessment. 8 Which worlds are relevant for evaluating a given assertion for truth will thus depend on parameters of the context of use/context of assessment pair. Consider the sentence:

(1)

There will be a sea battle tomorrow

as uttered by me at t1. Whether (1), as uttered by me at t1, is true depends partially on the context of assessment. My current context of use and my current context of assessment determine a set of overlapping worlds that all run through the current time t. Since a sea battle is taking place in some of these worlds but not in others, it is indeterminate with respect to these worlds whether there will be a sea battle tomorrow. So, (1), as uttered by me at t1, is neither determinately true nor determinately false. 9 But suppose now that you are evaluating (1), as uttered by me at t1, at some time t2 late tomorrow. My context of use and your context of assessment together determine a set of overlapping worlds that run through the times determined by my context of use and your context of assessment. Relative to this set, there is a determinate answer to the question of whether a seabattle is taking place the day after my assertion at t1. So, relative to this set, (1), as uttered by me

‘Conditionals and Relativism’, manuscript, Paul Bohgossian, ‘What is Relativism’, manuscript, Crispin Wright, ‘True Relativism’, manuscript. 7 See Nuel Belnap and Mitchell Green, ‘Indeterminism and the thin red line’, Philosophical Perspectives 8 (1994), pp. 365-388, and Nuel Belnap, Michael Perloff, and Ming Xu, Facing the Future: Agents and Choices in Our Indeterministic World (Oxford: Oxford University Press, 2001). 8 If the universe is truly indeterministic, there is no unique world that runs through the moments determined by the context of use and the context of assessment. 9 See ‘Truth in the Garden of Forking Paths’, §3.

4

at t1, has a determinate truth-value. Relativism is thus able to accommodate both of our intuitions about future contingents.

III. Traditional Semantics and Future Contingents MacFarlane’s relativistic semantics solves the problem of future contingents by relativizing sentence truth to a context of use and a context of assessment. But it is a very radical view. It requires us to revise the assumption of standard semantics that sentence truth is relative only to a context of use. Still, if no competing view is able to accommodate all our intuitions about future contingents, perhaps it is time to reconsider the standard assumptions of traditional semantics. I think, however, that there is no need to relativize sentence truth to contexts of assessment, as standard semantics (plus an account of truth-value gaps) can accommodate all our intuitions about future contingents. The alleged problem for standard semantics is that it cannot account for the intuition that from the perspective of an evaluator located some time late tomorrow, the sentence:

(1)

There will be a sea battle tomorrow

as uttered by me at t1, has a determinate truth-value. Suppose, for instance, that you and I are standing in the harbor at t2 in the midst of a sea battle. You might say: 10

10

I am aware that we wouldn’t ordinarily talk this way. However, I will begin with the more difficult case, and then turn to an easier (and more natural) case below. Also, I originally had you saying ‘the sentence “there will be a sea battle tomorrow”, as uttered by you yesterday, turned out to be true’. However, as an anonymous referee pointed out to me, this weaker version of the puzzle emerges on the growing block conception of Tooley, and on the shrinking tree model of McCall (which MacFarlane rejects). As the referee further notes, the stronger version of the puzzle, now presented in the paper, does not emerge on the growing block or shrinking tree conceptions. The growing blocker posits new facts which change the truthvalue of the uttered sentence from neither to either true or false, while McCall posits branch attrition which allows us to say why (2) holds despite (1).

5

(2)

The sentence ‘there will be a sea battle tomorrow’, as uttered by you yesterday, was true at the time of utterance.

The intuition that by asserting (2) you have said something true is very strong. In fact, it would be extremely odd for you to deny at t2 that ‘there will be a sea battle tomorrow’, as uttered by me at t1, was true at the time of utterance. The problem for standard semantics then is that (2), as uttered by you at t2, seems to be attributing the property was true to a sentence that was asserted, or uttered, by me at t1. So, whether or not this attribution is correct, it seems, will depend on, and only on, the truth-value of the sentence, as asserted by me the day before. Or so the standard objection goes. This standard objection, however, betrays a mistaken view of direct speech reports. The reported sentence in (2) is supposed to adequately represent the sentence I uttered yesterday: it is merely mentioned, not used, in my current context. As François Recanati has argued, 11 however, there is good reason to think that reported sentences are used, not simply mentioned. One reason is that the quoted material in a direct speech report may be available for copying, as in: 12

(3)

‘I’m going to talk to the doctor’, she said; and she did.

‘And she did’ is elliptical for ‘and she did talk to the doctor’. The elided material ‘talk to the doctor’ is available for copying. But this suggests that the quoted material is used. For if the sentence were merely mentioned, the quoted material would not be available for copying, witness ‘ “I’m going to talk to the doctor” is a well-formed sentence; and she did’.

11

‘Indexicality and context-shift’, Conference Paper, Workshop on Indexicals, speech acts, and logophors, Harvard University, (2004, 11-20), § 3.2. 12 The first argument is Recanati’s (§ 3.2, example 4); the second is my own.

6

Another reason to think that a reported sentence in a direct report is used not simply mentioned is that expressions in the embedding matrix clause can depend anaphorically on expressions in the reported sentence. Consider, for instance:

(4)

‘Give me your moneyi, or I’ll shoot’, he said, but I didn’t give iti to him.

The pronoun ‘it’ in the matrix clause is anaphoric on ‘your money’. But this requires that ‘your money’ picks out an individual for ‘it’ to refer to. Recanati suggests that direct speech reports create a shifted context that determines the semantic values of the indexicals in the report. So, in my assertion of ‘and then John said, “I am leaving now” ’ the semantic value of ‘I’ is John, not me, and the semantic value of ‘now’ is a time in the past, not the current time. Direct speech reports can thus be seen to function as contextshifters, or monsters, as Kaplan called them. 13 Kaplan thought there weren’t any context-shifting operators in English, but didn’t rule out direct speech report shifts. 14 Given Recanati’s account of direct speech reports, we can provide a more adequate account of post-sea battle reports. When you assert:

(2)

The sentence ‘there will be a sea battle tomorrow’, as uttered by you yesterday, was true at the time of utterance.

at t2, your context of use fixes the semantic values of the indexicals in the matrix of (2). So, it fixes the semantic values of ‘you’ and ‘yesterday’. But the direct speech report in (2) creates a 13

D. Kaplan, ‘Demonstratives’, at p. 510. Kaplan: ‘no operator can control the character of the indexical within its scope, because they will simply leap out of its scope to the front of the operator. I am not saying we could not construct a language with such operators, just that English is not one. And such operators could not be added to it’. 14 As Kaplan puts it, ‘there is a way to control an indexical, to keep it from taking primary scope, and even to refer it to another context. Use quotation marks. If we mention the indexical rather than use it, we can, of course, operate directly on it’, at pp. 510-11.

7

shifted context. So, the context with respect to which the quoted sentence is interpreted is not your context of use but a context in which I am the speaker and the time is some time yesterday. Any indexicals in the scope of the direct speech report are assigned semantic values relative to this shifted context. ‘Tomorrow’, for example, is assigned the day of t2. So with respect to the shifted context the quoted sentence expresses the proposition that it will be the case that a sea battle takes place on the day of t2. Besides fixing the semantic values of the indexicals that occur in the matrix of (2), your context of use at t2 also determines an index with respect to which the content of (2), relative to your context at t2, is evaluated. The content of the quoted sentence in (2) is not evaluated with respect to the index determined by your context of use at t2. But the reason for this is not that it is quoted. For direct speech reports do not by themselves shift any of the parameters of the index determined by the context in which the report is made. To shift the index we need an indexshifting operator. For example, the proposition expressed by ‘I am hungry’ is evaluated with respect to the index determined by the context of use in an assertion of ‘John says, “I am hungry” ’ but with respect to a shifted index in an assertion of ‘John said, “I am hungry” ’. The reason the quoted sentence in (2) is not evaluated with respect to the index determined by your context at t2 is that the matrix of (2) contains a past tense operator (‘… was true at the time of utterance’). The past tense operator shifts the time feature from the time of speech to some time before the time of speech. So, the proposition that there will be a sea battle on the day of t2, expressed by the quoted sentence in (2), is evaluated with respect to a time that precedes t2. But the world feature of the index is not shifted. For the past tense operator is selective: it operates only on the time feature of the index, not the world feature. So the proposition that there will be a sea battle on the day of t2, expressed by the quoted sentence in (2), is evaluated with respect to an index that contains a time t1 that is earlier than the day of t2, and a set of overlapping worlds determined by your context at t2. Since all the overlapping worlds determined by your context at t2 contain a sea-battle at t2, the proposition expressed by the quoted

8

sentence in (2), viz. the proposition that there will be a sea battle on the day of t2, is true; hence, (2), as uttered by you at t2, is true. We can thus account for the truth of (2) without relativizing sentence truth to contexts of assessment. (1) is neither determinately true nor determinately false with respect to my context of use at t1. Nonetheless, when you report the sentence I uttered, the proposition expressed by the reported sentence relative to the shifted context is evaluated with respect to an index that is determined by your context of use and index-shifting operators. Since the matrix of (2) does not contain any operators on the world parameter of the default index determined by your context at t2, the proposition expressed by the reported sentence in (2), viz. the proposition that there will be a sea battle on the day of t2, was true. In the above cases we took you to utter the very same sentence that I uttered yesterday. But evaluators may not actually utter the reported sentences; they may just quietly assess them for truth. However, such cases can be treated in the same way as (2) as well; for, on Kaplan’s theory, any sequence of parameters that includes a speaker, an addressee, a set of worlds, a time, and a location counts as a context. The speaker need not be speaking, and the addressee need not be listening. So, a context in which someone quietly assesses a sentence for truth is a context. A sentence assessed in such a context can be treated as reported, and so can be treated in the same way as the sentence reported in (2). Standard semantics (plus an account of truth-value gaps) can thus accommodate the intuition that when you assert ‘the sentence “there will be a sea battle tomorrow”, as uttered by you yesterday, was true’ at t2, the proposition expressed by the reported sentence has a determinate truth-value. There is thus no need to relativize sentence truth to a context of use and a context of assessment. Granted, (2) is not the most natural way of evaluating past assertions. A more natural way of evaluating (1), as uttered by me yesterday, would be to say something like ‘what you said

9

yesterday was true when you said it’. But, as MacFarlane has recently argued, 15 such cases are unproblematic for traditional semantics. ‘What you said yesterday was true when you said it’ is evaluated with respect to the index determined by the speaker’s context. The past tense (‘… was when you said it’) shifts the time feature of the index determined by the speaker’s context, not the world feature. As a sea battle is taking place at every world in the index, the proposition denoted by ‘what you said yesterday’ is true at the index.

IV. ‘Actually’ MacFarlane admits (though for different reasons) that it may be that traditional semantics can perhaps account for all of our intuitions about past assertions of future contingents. 16 However, he says, traditional semantics cannot account for future contingents that contain the word ‘actually’. 17 Suppose, for instance, that I assert the following sentence at t1:

(5)

There will actually be a sea battle tomorrow

According to MacFarlane, (5) makes trouble for standard semantics. For suppose that you assert the following sentence at t2 (in the midst of a smoking sea battle):

(6)

The sentence ‘there will actually be a sea battle tomorrow’, as uttered by you yesterday, was true at the time of utterance.

15

‘Truth in the Garden of Forking Paths’. See also Richard Heck, Jr., ‘MacFarlane on Relative Truth’, in E. Sosa and E. Villanueva (eds.), Philosophical Issues 16 (2006): Philosophy of Language, pp. 88-100, who offers different arguments for thinking that standard semantics can explain the shifty nature of utterances of sentences such as ‘what Bill said was true (when he said it)’. 16 ‘Truth in the Garden of Forking Paths’. 17 The following is a re-write of MacFarlane’s original objection. The basic idea underlying the original objection is the same.

10

Given standard semantics, (6) is true at your context of use at t2 iff the proposition that there will be a sea battle at t2 is true at t1 and the set of overlapping worlds that run through t1. The reason is that, given the standard treatment of ‘actually’, ‘actually’ as it occurs in ‘there will actually be a sea battle tomorrow’ always takes us from the worlds of evaluation to the worlds of speech. As MacFarlane puts it: No matter how deeply embedded we are, no matter how far the world of evaluation has been shifted, the actuality operator returns it to the world of the context of use … the effect of adding an actuality operator to the front of the a sentence is to shift the world of evaluation to the world of [the context of use]. 18 If the standard treatment of ‘actually’ is correct, then (6) is true relative to your context of use at t2 iff the proposition that there will be a sea battle at t2 is true at t1 and the set of overlapping worlds that run through t1. But the proposition that there will be a sea battle at t2 is neither determinately true nor determinate false with respect to this time and this set of worlds (as some of the worlds contain a sea-battle while others don’t). So, given standard semantics, (6) is false with respect to your context of use at t2. But intuitively, it is true. I think, however, standard semantics isn’t committed to any such thing. For it is not clear that ‘actually’ plays any significant semantic role in ordinary English. If I say ‘actually, there will be a sea battle tomorrow’ in reply to your allegation ‘tomorrow will be another boring day’ I am merely using ‘actually’ to indicate that I disagree with you. Another example that illustrates the same point: John, an inhabitant of the actual world with special powers, is communicating with Mary, an inhabitant of a nearby world w1. In w1 the inhabitants are approaching a presidential election.

Mary: I wonder who is going to win. A democrat I hope. John: No, a republican will actually win the election at your world. 18

‘Truth in the Garden of Forking Paths’, § 10, at p. 23.

11

John clearly doesn’t intend to say that at Mary’s world, a republican will win the election in the actual world. ‘Actually’ has other non-semantic uses. As Jason Stanley points out, sometimes ‘actually’ simply indicates ‘a “colouring”, or “tone”, like the difference between typical unphilosophical usages of “truth” and “absolute truth” ’. We might use “the actual president” (as in “the ACTUAL president of the United States came by for a visit”), ‘not to distinguish different possibilities, but rather for emphasis, to convey [our] surprise’. 19 MacFarlane is right that standard semantics is in trouble if the role of ‘actually’ is to shift the world(s) of evaluation to the world(s) of speech. If it functions in this way, then it is the modal equivalent of ‘now’, which shifts the time of evaluation to the time of speech. ‘It will be the case that I am happy that I am writing now’, for example, does not mean that it will be the case that I am happy that I am writing in the future. Rather, it means that it will be the case that I am happy that I am writing at the time of speech. Because ‘now’ shifts the time of evaluation to the time of speech, it does not admit of disquotation. The following claim, for example, is less than obviously true: 20

(7)

If John spoke truly when he said ‘I am hungry now’, then he is hungry now.

If MacFarlane is right that ‘actually’ is the modal equivalent of ‘now’, then we should expect it to behave in the same way as ‘now’. Thus, we should not expect the following claim to be obviously true: 19

See J. Stanley, ‘Rigidity and Content’, in Logic, Language and Reality: Essays in Honor of Michael Dummett, R. Heck (ed.), (Oxford University Press, Oxford, 1997), pp. 131-156, at p. 137. Stanley thinks ‘actual’ makes a difference to the semantic content of sentences in which it occurs, even if it does not make a difference to the assertoric content. But the further examples presented suggest that ‘actual’ doesn’t even make a difference to the semantic content of the sentence. 20 H. Cappelen and E. Lepore, ‘Context-Shifting Arguments’, in Hawthorne and Zimmerman (eds.), Philosophical Perspectives 17: Language and Philosophical Linguistics, (2003), pp. 25-50.

12

(8)

If John had spoken truly when he said ‘I am actually hungry’, then John would actually have been hungry.

For if the content of ‘actually’ returns us to the world of speech, then reported occurrence of ‘actually’ would take us to a non-actual world, whereas the non-reported occurrence of ‘actually’ in the consequent would take us to the actual world. So, the counterfactual in (8) wouldn’t be obviously true. But as (8) seems obviously true, there is good reason to think that ordinary language uses of ‘actually’ do not function in the way suggested by MacFarlane. Of course, there is also a technical use of ‘actually’. On this use, ‘actually’ functions in exactly the way suggested by MacFarlane. So, if the reported occurrence of ‘actually’ in

(6)

The sentence ‘there will actually be a sea battle tomorrow’, as uttered by you yesterday, was true at the time of utterance.

is interpreted in this way, then (6) is false. But I doubt that this poses any special problem for standard semantics. For it is not clear that we have strong intuitions about the truth-value of sentences with highly technical terms. It may well be that the feeling that (6) is true owes to the fact that the quoted sentence would have had a determinate truth-value if we had used ‘actually’ in one of its non-technical senses. When given a technical interpretation, cases like (6) show that standard semantics and relativistic semantics come apart. But they do not show that standard semantics gets things wrong.

V. Conclusion Prima facie, it seems that ‘there will be a sea battle tomorrow’ is neither determinately true nor determinately false with respect to our current context of use, but “becomes” true tomorrow if it

13

turns out that a sea battle is indeed taking place. Because traditional semantics takes sentence truth to be relative only to a context of use, it supposedly cannot accommodate this appearance. The problem of future contingent thus seems to motivate a shift from traditional semantics to a semantics that rejects the assumption that sentence truth is relative only to a context of use. On MacFarlane’s relativistic semantics, sentence truth is relative to a context of use and a context of assessment. So, the sentence ‘there will be a sea battle tomorrow’ is neither true nor false relative to my current context of use and my current context of assertion, but it is true relative to my current context of use and the context of someone who evaluates it tomorrow when a sea battle is indeed taking place. I have argued, however, that despite initial appearances to the contrary, traditional semantics can accommodate all of our intuitions about future contingents. Suppose that yesterday you said ‘there will be a sea battle tomorrow’. Today I am evaluating the sentence ‘there will be a sea battle tomorrow’ with respect to your context of use yesterday. On an adequate account of direct speech reports, the reported sentence in ‘the sentence “there will be a sea battle tomorrow”, as uttered by you yesterday, was true at the time of utterance’ is used in addition to being mentioned. The reported sentence is interpreted with respect to a shifted context created by the direct speech report, and the proposition it expresses is evaluated with respect to an index with a shifted time parameter. Since the past tense operator does not shift the set of worlds at which the content is evaluated, the reported sentence is true relative to the reporter’s context of use. The truth-predicate must thus be treated differently when it is restricted to the meta-linguistic level and when it is not so restricted. 21 Here is my suggestion in a nutshell: object-language occurrences of both S and ‘S is true’ are true iff S expresses a proposition that is true at the index determined by the context of the 21

Compare ‘yesterday when S said “there will be a sea battle tomorrow”, S was speaking the truth’ vs. ‘yesterday when S said “there will be a sea battle tomorrow” what S said wasn’t yet determinately true’. If what I am saying is right, then we should expect ordinary speakers to make only the first sort of metalinguistic claim. Thanks to André Gallois, Carrie Jenkins, Peter Ludlow, Robbie Williams, an audience at the Pacific Meeting of the APA in San Francisco, 2007, and an anonymous referee.

14

speaker who uttered S or ‘S is true’. But the truth-predicate as restricted to the meta-linguistic level is the usual one (given Kaplan-style semantics). So, in the meta-language, S as uttered in c is true simpliciter iff the proposition expressed by S in c is true at the index determined by the context of the speaker who uttered S (and not the context of the semanticist). I conclude that the problem of future contingents does not motivate a shift from nonrelativistic semantics to relativism. 22

22

I am especially indebted to Kent Bach, David Chalmers, André Gallois, John Gabriel, John Greco, John Hawthorne, Richard Heck, Mark Heller, David Jehle, Carrie Jenkins, Max Kölbel, Ernie Lepore, Ishani Maitra, Julien Murzi, Stephen Neale, Joe Salerno, Eric Schliesser, Walter Sinnott-Armstrong, Jason Stanley, Brian Weatherson, Ralph Wedgwood and Robbie Williams for discussion of these and related issues and to Katherine Hawley, Peter Ludlow, John MacFarlane and two anonymous referees for comments which improved the paper considerably. The paper was presented at the Pacific Meeting of the APA, San Francisco, 2007. I am grateful to the audience and my commentator Peter Ludlow for their wonderful comments, suggestions and encouragements.

15